THE BENTO BOX REVEAL
If you’ve been watching the AI news this week, you probably saw the massive 500,000-line code leak from Anthropic’s Claude agent.
When you look at it, you realize something pretty validating...
Even the biggest AI companies in the world have to build massive, complex mazes of rules just to keep their AI from hallucinating or drifting off-topic in a conversation!
I know we've all felt a smaller version of that exhaustion.
You dump your messy thoughts, some background context, and a bunch of raw data into a chat box... and then you spend the next 20 minutes arguing with the AI to get it to format things correctly. 🤣
Recently, I shifted how I work to fix this...
And I want to share the tool I built to help you do the same.
I stopped writing conversational paragraphs and started organizing my prompts into strict, isolated compartments.
I call it Bento Box Prompting.
Instead of a messy pile, a Bento Box prompt looks like this:
[CONTEXT] The background info.
[RULES] Exactly how it must behave.
[INPUT] The raw data you need processed.
When you isolate the information this way, the physics of the machine actually change.
The AI stops trying to act like a chatbot guessing your intent, and simply fills the logical space you provided.
It becomes a deterministic machine...
To make this frictionless for you, I built a custom Bento Box Builder Gem.
Take whatever prompts you currently like using...
Drop them into this "Bento Box Prompt Builder" Gem...
It will automatically engineer them into a perfect Bento Box structure for you to use.
👉 https://bento.automatedwith.tech
Test it out today.
I promise you will immediately feel the shift from being an "AI User" to an "LLM Operator."
(Side note for the developers and tinkerers in the group: If you want to see how we use these simple Bento Boxes to build autonomous digital workers and APIs without writing complex Python, I just updated our open-source L.E.A.P. templates on the GitHub repo. You can grab those here: https://github.com/tomi-mercer/leap-design-pattern/ )
Give the Gem a spin on a task you are working on today...
Drop a comment below and let me know how the output compares to your usual prompts!
I just tried it for an email I was sending out.
Yesterday, I have an AI that writes email through N8N with no context. I have to give it the context. I asked it to write an email, summarizing our last weeks content and previewing our next weeks content. I gave it YouTube links to be able to develop that content.
The result was absolute garbage.
Then I used an LLM that I normally use and gave it the same prompt.
Result: more garbage.
Then I argued with it.
Then I argued with it some more.
Then I changed the setting to Deep Research and finally got something that was a version of what I was looking for.
Today, I put my original prompt into the Bento Box Prompting Gemini.
I then put the output of that prompt into the N8N email writer.
It gave me an email that exceeded my expectations on the first try.
Michael Boyle That is freaking awesome! Nice win.
Michael Boyle Thanks for dropping your bento success!!
There's an even more refined version of BENTO on the way (and you'll be seeing the Gem adjust accordingly over the next week or so).
Should help you (and everyone else!) take things to a whole new level!